Triple Space‐Polarization On–Off‐Key Modulation With AlN‐Based Tri‐Focal Metalens Authors: Zhi‐Yan Lin, Chen‐Yi Tsai, Bo‐Yi Chen, Huang‐Yu Wei, Hsiao‐Hsuan Yang, Yuh‐Jen Cheng, Hideto Miyake, Chia‐Yen Huang Published: 2026 ABSTRACT ABSTRACTWe designed and demonstrated a tri‐focal metalens (TF‐lens) made of high‐temperature‐annealed AlN (HT‐AlN) on sapphire. The HTA‐AlN film possesses an ultrawide bandgap (E g = 6.2 eV), atomic‐level smoothness, and single‐crystalline quality, making it an exceptional candidate for a metasurface in visible spectral wavelengths. By incorporating polarization‐insensitive zone‐plate null rings into a polarization‐selective bifocal metalens, the incident light will be focused into two or three foci on the focal plane, each with a distinct polarization state. The extinction ratio between the on and off peaks exceeds 1000 after subtracting the background signal, and the absolute value of S 3 is measured to be higher than 0.96 at the side foci. By manipulating the incident polarization state and the angle of the linear analyzer behind the TF‐lens, all combinations of the on–off keying at three distinct foci can be generated. Introduction Compound semiconductors are strong candidates for metasurfaces due to their high refractive index and high transparency in the subbandgap spectral region. A compound semiconductor‐based metasurface can also be monolithically integrated with active optoelectronic devices made of the same materials, such as light‐emitting diodes [1, 2, 3, 4, 5] and surface‐emitting lasers (SEL) [6, 7, 8, 9, 10, 11], to modulate the out‐coupling beam within an ultra‐compact space. For example, the beam from the SELs can be split into two beams with orthogonal polarizations by modulating the Pancharatnam–Berry (geometric) phase of birefringent meta‐atoms [12]. A particular optical spin can be extracted from a linearly polarized light by meta‐atoms with mixed phases [13, 14]. A beam with orbital angular momentum (OAM) can also be generated by imposing a helical phase profile encoded in the metasurface [15, 16, 17]. The generation of optical spin and OAM enhances the information density of photons with the same wavelength, which is critical for next‐generation information technologies, such as quantum information [18, 19, 20] and visible‐light communications [21]. Among all the compound semiconductors, AlN has the widest wavelength compatibility for metasurfaces due to its ultrawide bandgap (E g = 6.2 eV) and a sufficiently high refractive index. Robust and efficient AlN metalenses have recently been demonstrated from 266 to 375 nm [22, 23, 24]. However, the polycrystalline nature of the conventional sputtered AlN film imposed practical challenges in the fabrication of AlN metasurfaces. As the dimensions of the meta‐atom are close to the grain size of the AlN nano‐column, fabricating the nano‐pillar with high pattern fidelity via dry etching becomes critical. Therefore, a homogeneous AlN film is strongly desired for sophisticated metasurfaces. Over the past decade, atomically flat and nearly single‐crystalline high‐temperature‐annealed AlN (HTA‐AlN) films on sapphire substrates have been developed as a growth template for efficient deep‐UV light‐emitting diodes [25, 26, 27]. These HTA‐AlN films are also an ideal platform for metasurfaces in the UV–visible spectral range. Figure 1a,b show the optical constants and surface morphology of the HTA‐AlN film used in this study. The refractive indices are above 2.2 from 200 to 700 nm, and the transmittance is above 50% down to 210 nm. The bi‐layer atomic‐terrace morphologies implied exceptional surface flatness and thickness homogeneity, which GaN on sapphire heteroepitaxy could not achieve at submicrometer thicknesses. The high film quality in submicrometer thickness is important for the meta‐atom's pattern fidelity. Because the sapphire substrate serves as the natural etch‐stop layer for III‐nitrides, the etching time window for meta‐atom fabrication is broader if the meta‐atom height matches the initial film thickness. As long as the meta‐atom library can cover 2π phases and has a decent transmittance, a lower pillar height is preferred from a fabrication perspective. Therefore, the thinner HTA‐AlN buffer could be a more favorable choice for a metasurface than GaN heteroepitaxy in the visible spectral region, even though GaN might have a higher refractive index. (a) The index dispersion and transmittance curve and (b) the 1 μm × 1 μm atomic‐force microscopy image of an HTA‐AlN on sapphire. In this work, we design and fabricate a tri‐focal metalens (TF‐lens) with HTA‐AlN as shown in Figure 1. By integrating the polarization‐sensitive bifocal lens (BF‐lens) and the polarization‐insensitive zone plate (ZP), the incident beam is split and focused into three space divisions with different polarization states. The on and off states at each division can be modulated by controlling the polarization of the incoming and outgoing beams using external polarization optics. The TF‐lens holds strong potential to enhance the capacity of applications that rely on the binary information transmitted by photons. Design and Simulation Construction of the TF‐Lens To independently manipulate photons with opposite spin states using a single array of densely packed meta‐atoms, both the propagation phase and geometric phase of polarization‐sensitive meta‐atoms must be engineered to fully exploit the available degrees of freedom. A schematic of the AlN rectangular nanofin structure is presented in Figure 2a. A hexagonal lattice is employed due to its higher filling factor, which helps mitigate efficiency degradation associated with the Nyquist limit. The lattice period is set to 360 nm, and the nanofin height is 600 nm. The Jones matrix presentation of light passing through the nanofin with the basis of the unit left‐circularly polarized (LCP) and right‐circularly polarized (RCP) spins is written as Equation (1) [13, 23]: where θ is the rotation angle between the reference axes and the principal axes of the ordinary ray and the extraordinary ray. The parameters To and Te represent the real‐valued electric‐field transmissivity, whereas φo and φe denote the associated propagation phases of the o‐ray and e‐ray, respectively. The To/Te and φo/φe of a nanofin with given Wx and Wy are calculated with Reticolo‐RCWA by assuming the polarization of the plane wave to be along the reference x‐axis. To further simplify the matrix formulation, we introduce the root‐mean‐square transmissivity (Trms) and the half‐wave plate factor (ηHWP), as defined in Equations (2) and (3): (a) The schematic structure of the AlN meta‐atom, (b) the color mapping of ηHWP2 with W x and W y , (c) the look‐up table of W x and W y for an optimized ηHWP2 value, and (d) schematic TF‐lens with F = 16 μm and a diameter of 10 μm. Substituting these into Equation (1) yields Equation (4): ϕu and ϕc are the effective propagation phases of the unconverted and converted spins, given by Equations (5) and (6): If the metasurface is constructed with meta‐atoms with ηHWP∼1, that is, the HWP ensemble, Equation (4) reduces to Equation (7): According to Equation (7), the LCP component of an arbitrary incident beam sees φL=φc+2θ, and the RCP component sees φR=φc−2θ; then, the polarization switches to the opposite spin. As long as the phase φc of the HWP ensemble spans a full 2π range, the ϕL and ϕR can be independently assigned. Figure 2b maps the calculated ηHWP2 as a function of W x /W y in Figure 2a. After filtering those with ηHWP2 > 0.99, ϕc as a function of Wx/W y is plotted in Figure 2c as the look‐up table for metasurface construction. The geometric parameters W x and W y are determined from Figure 2c based on a target phase φc=φL+φR∕2. Equations (8) and (9) are the assigned metalens profiles for the RCP and LCP components, respectively. The focal length (F) is identical, but the foci are spatially displaced along the x‐axis from the origin: where F = 16 μm and x 0 = 3 μm for the following simulations. The third focus is introduced by incorporating the zone plate into the bifocal metalens. The radius of each ring is defined by Equation (10): The regions with r2m−2 1000 after the background noise subtraction, implying negligible optical crosstalk for practical applications. Notably, the I s /I c ratios in Table 1 are natural outcomes of the TF‐lens following the construction rules in Section 2.1, and those values could change with a different given NA. There might be a target I s /I c ratio of the TF‐lens for some applications. In the following sections, we will discuss the targeted ratio and how it can be actively achieved regardless of the given NA. Triple Space‐Division On–Off Keying (3SD‐OOK) In many applications, such as photolithography and data communication, photons carry only binary information: on (= 1) and off (= 0). If the TF‐lens can be used to transmit on and off states at three different points, the information density per shot is tripled by spatial‐division multiplexing. To implement all possible combinations of the triple‐spatial‐division on–off keying (OOK), a quarter‐wave plate (QWP2) and a linear polarizer (LP) are integrated behind the objective lens, as illustrated in Figure 6a. By leveraging Malus's law, the relative intensity between the focal points can be precisely modulated by attaching polarization optics to achieve seven distinct OOK states as summarized in Table 2. (a) The schematic setup for the on–off keying generation at each focus and (b) the measured intensity profile on the final focal plane based on the OOK state marked at the corner and configuration listed in Table 2. TABLE 2Summary of the QWP1 and LP configurations and intensity ratios for all OOK states except the 000.Polarization state I 1:I 2:I 3 w/o LP β of LP I 1:I 2:I 3 w/LPOOK stateAfter QWP1w/o QWP2w/QWP2LP‐y E1∥Rˆ E2∥yˆ E3∥Lˆ E1∥xˆ+yˆ E2∥yˆ E3∥xˆ−yˆ α:1:α 0 α/2:1:α/2111 π/2 1:0:1101LCP E1∥Rˆ E2∥Lˆ E1∥xˆ+yˆ E2∥xˆ−yˆ 2α:1:0 π/4 1:0:0100 β′ 1:1:0110 −π/4 0:1:0010RCP E2∥Rˆ E3∥Lˆ E2∥xˆ+yˆ E3∥xˆ−yˆ 0:1:2α −β′ 0:1:1011 −π/4 −π/4 001 Note: 0 = off, 1 = on. Summary of the QWP1 and LP configurations and intensity ratios for all OOK states except the 000. E1∥Rˆ E2∥yˆ E3∥Lˆ E1∥xˆ+yˆ E2∥yˆ E3∥xˆ−yˆ E1∥Rˆ E2∥Lˆ E1∥xˆ+yˆ E2∥xˆ−yˆ E2∥Rˆ E3∥Lˆ E2∥xˆ+yˆ E3∥xˆ−yˆ Note: 0 = off, 1 = on. In the first configuration, the fast axis of QWP1 is aligned with the LP‐y incident light to preserve its original polarization state as it enters the TF‐lens, thereby forming three focal points. After magnification by the objective lens, the light passes through QWP2 (fixed with its fast axis along the y‐axis), which converts the circular polarization of the foci into linear states along the xˆ±yˆ direction. Assuming I s /I c to be α under LP‐y incidence, the intensity ratio among three foci after passing through the LP at β=0 becomes α/2:1:α/2 by Malus's law. Therefore, if we want to transmit the 111 state with uniform intensity across three points, α must be 2. In this report, the α of the fabricated device is 0.71, as noted in Table 1. The strategy of actively tuning α will be discussed later in Section 3.3. If the LP is rotated to β=π/2, the central channel is blocked, thereby generating the 101 OOK state. Subsequently, if the TF‐lens is under pure circularly polarized incidence (LCP and RCP), the TF‐lens generates only two foci. The pristine intensity ratios without LP will be 2α:1:0 or 0:1:2α. Because these configurations involve only two active foci with orthogonal polarization states, the 100, 010, and 001 OOK states can be selectively controlled by setting β=±π/4 to allow only a single beam to pass. For the 110 and 011 states, the required β to equalize the intensities of the two “on” states can also be derived via Malus's law, yielding β=±β′=±π4−atan12α. Finally, the 000 state is achieved by turning off the source. The experimentally captured CCD images of the seven OOK states, following the configurations in Table 2, are shown in Figure 6b. Notably, the absolute peak intensities among the seven combinations in Table 2 are not necessarily equal under the same incident power due to the conservation of energy. If the application requires uniform peak intensity across seven OOK states, an additional input‐power control unit is required. Potential Applications TF‐Lens as a Generalized Polarization Splitter From Table 1, I s /I c under LP‐y incidence is around α= 0.71. Considering a unit y‐polarized electric field, Ein = yˆ, field is incident; the electric field at the focal plane, EFP, can be written as in Equation (11): where E 0 is the normalizing amplitude, Lˆ and Rˆ are the unit LCP and RCP electric fields, and P˙i presents the ith focal point position. For an incident beam with an arbitrary polarization, Ein, the EFP can be generalized to Equation (12): where Lˆ,Ein is the inner product of Ein and Lˆ, same for Rˆ,Ein. If the electric field at each focus is collected by different optical fibers, the TF‐metalens serves as a generalized polarization beam splitter, extracting the LCP and RCP components from P 1 and P 3, respectively, while simultaneously keeping part of the original beam from P2. The α can be intentionally designed over a wide range, as depicted in Figure 7. If we fill the ZP rings with the meta‐atoms of the BF‐lens design, α increases with the BF‐lens's weight. On the other hand, if part of the BF‐lens meta‐atoms are replaced with polarization‐insensitive meta‐atoms with a π‐phase shift, α decreases with the increasing ZP's weight. Schematic representation of tuning α by partial replacement of BF‐lens with π‐shifters (left) or gap‐filling with BF meta‐atoms (right). The TF‐lens could be used to extract the relative phase of coherent photons, an important parameter in quantum optics [28, 29, 30]. For example, considering ideal photons under the Bloch sphere presentation, ψ〉=cosθ|Lˆ〉+sinθeiϕ|Rˆ〉 incident on a TF‐lens, we can analyze its relative phase, ϕ, by measuring the x‐polarized component intensities on the focal plane. The x‐component intensity at the P i , I ix , can be calculated by substituting ψ〉 into Equation (12), yielding Equations ((13), (14), (15)): Then, θ and ϕ can be evaluated from the measured I ix with a precalibrated α. The derivation of Equations ((13), (14), (15)) and the extraction of θ and ϕ can be found in Supporting Information S1. Compatibility With Other Materials Platforms Beam‐splitting metalenses have been demonstrated across various material platforms, and the objectives varied from the wavelength [31] and polarization states [23, 32, 33] to the topological charge of optical vortices [34, 35]. Some representative works are summarized in Table 3. For metalenses built on the simultaneous modulation of geometric and propagation phases (i.e., the hybrid phases), the number of foci is limited to 2 for a single‐phase profile due to the limited degrees of freedom. One could also increase the number of foci by imposing multiple‐phase profiles on the same plane by meta‐atom sampling or tiling the area with different metalenses. We refer to these approaches as spatial multiplexing. In this work, embedding concentric null rings can be considered a derivative of spatial multiplexing. However, our approach only removes meta‐atoms rather than replacing them with different ones. Therefore, it shall be more broadly compatible with existing hybrid‐phase‐based bifocal metalenses across other material platforms, thereby further enhancing their functionality. TABLE 3Summary of the beam‐splitting multifocal metalenses on the forming materials, number of foci, objectives, and approaches.MaterialsNumber of fociObjectiveApproachRef.AlN3PolarizationHybrid phases + zone plateThis workAlN2PolarizationHybrid phases[23]Ag2PolarizationSpatial multiplexing[32]Liquid crystal2PolarizationHybrid phases[33]TiO2 3WavelengthSpatial multiplexing[31]TiO2 4Topological chargeSpatial multiplexing[34]TiO2 2 × lTopological chargeAzimuthal quadratic phases[35] Summary of the beam‐splitting multifocal metalenses on the forming materials, number of foci, objectives, and approaches. Conclusion In conclusion, we designed and fabricated a tri‐focal metalens that integrates a polarization‐insensitive zone plate and a polarization‐sensitive bifocal metalens. The distinctive polarization states at three foci enable all possible combinations of the on and off states in each space division with a single laser pulse. We believe that the unique polarization response of the tri‐focal metalens will enhance capacity and functionality in many applications that rely on transmitting binary information with photons. Author Contributions Zhi‐Yan Lin: investigation, methodology, validation. Chen‐Yi Tsai: writing – original draft, formal analysis. Bo‐Yi Chen: writing – review and editing. Huang‐Yu Wei: visualization. Hsiao‐Hsuan Yang: software, visualization. Yuh‐Jen Cheng: visualization, supervision. Hideto Miyake: resources. Chia‐Yen Huang: conceptualization, writing – original draft, writing – review and editing, project administration, methodology. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Supporting Information S1 Acknowledgments This work was financially supported by the National Science and Technology Council (NTSC) of Taiwan under Contract No. NTSC‐114‐2923‐E‐A49‐015‐MY2 and NSTC‐114‐2221‐E‐A49‐040‐MY3. Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request. References Associated Data Supplementary Materials Supporting Information S1 Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.